This report shows that the Satisficing Trade-Off Method (STOM), one of interactive multi-objective programming techniques, can be effectively applied to livestock ration formulation. It is recognized that the proposed method makes ration formulation very easily, rapidly and flexibly. The method is characterized as follows: 1) Multi-objective problem is simply formulated as a linear programming problem by using the aspiration level approach. 2) By adjusting the aspiration levels of nutritional requirements and/or feed cost, it is very easy for decision makers to set flexibly their requirements and to get surely satisfactory solutions. 3) It is more speedy to get satisfactory solutions by using the automatic trade-off analysis with parametric optimization techniques in Liner Programming. 4) It is easy to deal with soft constraints, for which the aspiration level is not necessarily attained, such as quantity of feed ingredients by treating the fuzzines of values of righthand sides of constraint function as the aspiration levels. 5) It is not necessary for decision makers to consider the role of objective and constraint fixed from the begining, because the soft constraint function may be interchangeable between objective function and hard constraint. 6) This method can be simultaneously applied to both a nutritional diagnosis and a ration formulation by setting actual intakes as soft constraints when the difference between the nutrient requirement and the actual intake must be made up by adjustments among feed ingredients.

View full abstract

In fields such as agriculture information system and agricultural system, artificial satellite image data have been used to obtain regional macro information. Especially in the region where the condition of land use is frequently altered, a classified map document containing condition of land use, regional agriculture plan etc., land cover based on remote sensing technique is needed. However it is very difficult to know the exact condition of land use because the change in land use is too fast and it takes too much time to make land use map with traditional mapping technique. Therefore, when the classified map of land cover is made by use of remote sensing image data, the current condition of land use is classified into several land cover items. However the classified items of land cover are determined based on the request of users, and they may not be suitable for the analyzed region. Moreover, even if they are suitable items, assessment of classification result may be different according to the sampling condition of supervised data (training area data) and samples' location of the data for assessing test area data, and assessment method of classification result and suitability of the data are probably important research theme. Till now "Equivocation Quantification" has been used for the assessment of classification result. However it is necessary to use crisp data for computing the equivocation quantification. When mixed patterns are included in one pixel, the different classification results can not be differentiated by the equivocation quantification. This paper introduces fuzzy theory to express the relationship of mixed pattern described above with membership function. And it is also proposed a new concept called "pseudo frequency", adapting the fuzzy data to compute equivocation quantification. It is also made clear in this paper that when classification result, accompanied with the making of principal map, is expressed by classification score, the equivocation quantification can be applied to fuzzy data.

View full abstract